The perception of shape from shading in a new light

Abstract

How do humans see three-dimensional shape based on two-dimensional shading? Much research has assumed that a 'light from above' bias solves the ambiguity of shape from shading. Counter to the 'light from above' bias, studies of Bayesian priors have found that such a bias can be swayed by other light cues. Despite the persuasive power of the Bayesian models, many new studies and books cite the original 'light from above' findings. Here I present a version of the Bayesian result that can be experienced. The perception of shape-from-shading was found here to be influenced by an external light source, even when the light was obstructed and did not directly illuminate a two-dimensional stimulus. The results imply that this effect is robust and not low-level in nature. The perception of shape from shading is not necessarily based on a hard-wired internal representation of lighting direction, but rather assesses the direction of lighting in the scene adaptively. Here, for the first time, is an experiential opportunity to see what the Bayesian models have supported all along.

Keywords: 3D shape perception; Illusion; Shape from shading; Vision; Visual perception.

Figure 1. Stimulus and results.

(A) An example stimulus. (B) Mean rating of perceived light direction versus the actual illumination direction in the room, as a function of experimental condition (Condition 1: Overhead Light; Condition 2: Light Below; Condition 3: Obstructed Light Below). 1.0 = light from above; 0.0 = light from below. Error bars are ±1 s.e.m.